THE STRUGGLE FOR PERFECTION

Although it is theoretically possible to write a program that perfectly predicts every possible situation that it might encounter, in practice that’s very difficult for nontrivial programs. For large applications, it is extremely difficult to plan for every eventuality. Errors in the program’s design and implementation can introduce bugs that give unexpected results. Users and corrupted databases may give the application values that it doesn’t expect.

Similarly, changing requirements over time may introduce data that the application was never intended to handle. The Y2K bug is a good example. When engineers wrote accounting, auto registration, financial, inventory, and other systems in the 1960s and 1970s, they never dreamed their programs would still be running in the year 2000. At the time, disk storage and memory were relatively expensive, so they stored years as 2-byte values (for example, 89 meant 1989). When the year 2000 rolled around, the applications couldn’t tell whether the value 01 meant the year 1901 or 2001. In one humorous case, an auto registration system started issuing horseless carriage license plates to new cars because it thought cars built in 00 must be antiques.

The Y2K problem wasn’t really a bug. It was a case of software used with data that wasn’t part of its original design.

This chapter explains different kinds of exceptional conditions that can arise in an application. These range from unplanned data (as in the Y2K problem) to bugs where the code is just plain wrong. With some advance planning, you can build a robust application that can keep running gracefully, even when the unexpected happens.